Skip to main content

How to Create an Interactive Avatar

J
Written by Jini(황진희)

Introduction

Interactive avatars can be used with or without creating a custom avatar.

DeepBrain AI allows users to either create their own avatar or use pre-built sample avatars for immediate use. Custom avatars provide a more personalized experience, while sample avatars are useful for quick testing and initial setup.


How to Create an Avatar

Users can create a custom avatar using the following methods:

  • Recording directly in the editor

  • Uploading a pre-recorded video

Once the recording is completed, the avatar enters an activation process before it becomes available for use in a project.

Only successfully activated avatars can be used for real-time conversation.


Recording Guidelines for Custom Avatars

To ensure high-quality interactive avatars, it is important to follow specific recording guidelines. The most critical factor is minimizing movement during both idle and speaking states.

Idle (Neutral State)

The idle state represents the avatar when it is not speaking.

During this phase:

  • Do not speak

  • Look directly at the camera

  • Avoid hand or body gestures

  • Only minimal natural movements such as breathing or blinking are allowed

A stable and calm posture is recommended. The goal is to maintain a natural but nearly motionless state.


Speaking (Talking State)

The speaking state represents the avatar while delivering responses.

During this phase:

  • Speak naturally on a single topic

  • Keep body and hands as still as possible

  • Avoid head movement and large gestures

  • Focus on clear speech rather than physical expression

Although the avatar is speaking, movement should remain minimal to ensure consistent output quality.


Key Recording Principles

For best results, the recording should follow these principles:

  • Movement should be minimized in both idle and speaking states

  • The avatar should maintain a stable, front-facing position

  • Excessive gestures or motion can reduce output quality

A good reference point is to maintain a level of stillness where the difference between idle and speaking is primarily in speech, not body movement.


Video Direction Guidelines

When creating a custom avatar video:

  • Avoid dynamic movements such as walking or large gestures

  • Focus on a stable, centered composition

  • Upper-body framing is recommended

  • Keep the visual consistent throughout the recording

The goal is to produce a video where the avatar appears steady and controlled, allowing the system to generate more accurate real-time responses.

Did this answer your question?